Rob Danielson wrote:
> Walt wrote:
>
>
>>ATRAC's bit depth reduction will occur mostly in those thousands of sampl=
es in
>>between each pulse, we don't need full bit depth to characterize silence
>
>
> I do believe this could be true from side by side experiences with MD
> and DAT using same mics, pres. For ambient recordings or those of a
> number of sound sources at a distance (not one, foregrounded call)
> there seems to be some acoustic detail that's lost with MD/ATRAC
> scheme. For example, long reverb tails that aren't there on the MD
> recording. When I try to coax more out of the low level MD recordings
> with careful eq, its evident there's more garbage than on the DAT.
> I'm also wondering if low saturation/saved bits could be part of
> Aaron's discrepancies. Saturation seems to be a critical variable,
> also the quality of D->A and monitoring people are using when making
> the judgements. Rob D.
What we are getting is a mishmash of different entire systems in the
samples being presented. Far too hard to sort out any commonality.
That's why I think it's important to work with a recording system for a
fair while with the same mics involved. And listen with the same set of
equipment. Only then will we develop a clearer picture as to what our
equipment can do.
What happens is that sometimes I've been absolutely certain I've found
something that the ATRAC hardware does, or somebody on this group or
elsewhere announces dogmatically how it will behave. Then a month, six
months, a year later have a chance to make a recording that should, if
whatever the theory is has validity, make the same effect. It simply
keeps not being so predictable. The requirement of reproducibility is
fundamental to science. A theory that cannot predict and be reproduced
is of very low value and soon discarded. This is a hardware circuit, it
should respond exactly the same way each time. It probably is, and
something else is making the difference.
As far as very low level parts of a recording, I don't think those have
been really examined with other types of recorders either. Have you
thought that it could just as easily be that the other recorders are
getting it wrong and the ATRAC is more correct? I'm hardly saying it is,
but you are starting from a biased mindset.
As far as long reverb tails not being there, I've seen plenty of times
when they were. That has not been consistent. I saw that proposed as a
theory years ago.
When I did the samples for the group blind test and other such
experiments with producing comparison samples that was a lesson in how
hard it is to analyze just one component of a working audio system, in
this case one chip on a crowded circuit board, which is connected to
others. Before jumping to the conclusion that anything you find is due
to the ATRAC encoding chip itself you have to eliminate all the other
variables. This is a real challenge. Not near as simple as the casual
samples being discussed.
If you want to say that a particular model MD recorder has thus and such
effect that's much easier and I have no problem with that, though I may
agree or not. But is far from the same thing as saying the ATRAC
encoding process itself is creating the effect. In fact it's different
from saying all MD recording systems have that effect.
I think we get into this sort of nit picking because of the mythology of
absolute perfection that's grown up about digital audio. It's not that
perfect in any of it's forms. It has a set of problems that are
different from those of analog, and it would be well if we got over the
starry eyed thing and got down into finding those and how we should live
with them. ATRAC is not alone and uncompressed does not mean perfection.
To only look at MD and ATRAC is missing most of the problems. Including
the big ones.
In the case of Aaraon's discrepancies, the sonogram reads to me that the
differences are of the most saturated parts, not the quiet parts. Since
they appear in the frequencies 300hz and below all the way down to the
bottom, they could be differences in the frequency bandpass
characteristics of the two machines. Or simply that the two machines end
up with a slight difference in levels. Or that different programmers
used slightly different methods to do the exact same thing, something
that's very, very common. It will be tricky to track down and no part of
the overall system involved can be ignored if it's to be tried.
Sound quality is a very subjective thing. Each person hears a sound
differently. The major component being their own brain in these
differences, and it's far from cold analysis there. Emotions, feelings,
attitudes all enter in. This makes it doubly hard to discuss these
things, we each hear a different sound. So the differences won't all
yield to analysis.
It's almost embarrassing to be associated with a group that puts up so
many varied problems with the immediate assumption that all are ATRAC
encoding errors, especially when it's not involved at all in some, and
often no other cause is even considered. It's pretty darned fuzzy
thinking, the group can do better. Take the blinders off.
Walt
________________________________________________________________________
________________________________________________________________________
|